Online Satellite Selection, Request Dispatching, and Service Provisioning in LEO Edge Constellations

Facing the growing demand for Low Earth Orbit (LEO) edge services, in order to address the manageability and economic issues in LEO edge computing, this study introduces an innovative two-timescale optimization approach designed to dynamically optimize satellite access selection, user request dispatching, and service replica placement. Integrating both online and offline optimizations, our method adapts to real-time fluctuations in user demand and satellite resources, effectively managing long-term decisions such as service migration and replica placement. We formalize this optimization challenge as a finite-horizon, integer-variable problem, taking into account both switching costs and resource utilization. Through extensive experimentation, our approach is proven to significantly balance performance enhancement and resource efficiency, and we prove the approximation ratio for each time slot the competitive ratio for the long-term cost. Our work contributes to the understanding of multi-timescale optimization in LEO edge computing and provides valuable insights for designing efficient control mechanisms in satellite-based systems.